| *** ralonsoh_ is now known as ralonsoh | 11:39 | |
| sean-k-mooney | gmaan: with the changes to rdo i wonder if it makes sense to contineu to use centos 10 stream going forward. i.e. if allma or rocky start providing the depencies we currently consume form rdo perhaps it would make sense to shift oru focus to rocky or alma as the primay rhel proxy | 14:49 |
|---|---|---|
| sean-k-mooney | if rdo is going to discontinue rpm support we should remove all usage of it form devstack perfering eple and alternitive souces for things like ovs, rabbitmq ectra. if distos like almalinux or rockey end up builing and delivering those it would seam prudent to leverage that effort going forward | 14:55 |
| sean-k-mooney | https://termbin.com/wb6t is a very quick ai survay of devstack current rdo deps | 14:57 |
| sean-k-mooney | gmaan: by the way we have moved ceph to contaiers as part fo the adoption fo cephadm | 15:11 |
| sean-k-mooney | how would you feel ablout doing the same for rabbit? if there isnt a suitable alternitive? | 15:11 |
| sean-k-mooney | eple has rabbitmq-server i belvie so just using eple woudl seam like the similar solution and usign the nvf sig repos for ovs/ovn | 15:14 |
| sean-k-mooney | but just barinstoming options before things break. | 15:14 |
| sean-k-mooney | actully trabbit upstream provides offical rpms https://www.rabbitmq.com/docs/install-rpm | 15:19 |
| sean-k-mooney | ah but only for 9* | 15:20 |
| mnaser | i think ovs is not part of rdo for quite sometime, that can be pulled in from the nfv sig | 15:20 |
| ykarel | rabbitmq also coming from messaging SIG | 15:21 |
| sean-k-mooney | rabbitmq-server x86_64 3.13.7-1.el10 centos10-rabbitmq | 15:27 |
| sean-k-mooney | we do still use rdo for ovs | 15:28 |
| sean-k-mooney | we dont need too | 15:29 |
| sean-k-mooney | https://zuul.opendev.org/t/openstack/build/c82282a5388b40a581abfd42fa9a1b7a/log/job-output.txt#8860 | 15:29 |
| sean-k-mooney | but we are installing | 15:29 |
| sean-k-mooney | rdo-openvswitch noarch 2:3.4-1.el10s delorean-master-testing 7.7 k | 15:29 |
| ykarel | yes the wrapper package is coming from RDO, actual binaries from NFV SIG | 15:29 |
| ykarel | for openvswitch | 15:29 |
| sean-k-mooney | ya | 15:29 |
| sean-k-mooney | so we may wantt o try disabling rdo entirly and see what breaks | 15:29 |
| sean-k-mooney | ykarel: is this somethign you have spent any time looking at | 15:31 |
| sean-k-mooney | it does seam we are mainly depeing on wrappers at this point | 15:32 |
| sean-k-mooney | outside of a few packages like liberasurecode-devel | 15:32 |
| sean-k-mooney | hum and uwsgi | 15:33 |
| ykarel | not checked much, but only few deps looks coming from RDO | 15:37 |
| mnaser | uwsgi could easily be pulle dform pip | 15:40 |
| sean-k-mooney | we avoid it becuase it fails to compile sometimes | 15:40 |
| sean-k-mooney | but it could | 15:40 |
| clarkb | it is particularly flaky on arm64. But even just building with -j >1 seems to make it really unstable | 16:08 |
| sean-k-mooney | we cloud just use coper or somethign to build that one package and pull that | 16:17 |
| sean-k-mooney | or build it as a binary wheel somewhere | 16:17 |
| sean-k-mooney | i was more just wodnerign what the blast radius of possible change might be if the rpms are notlogner built. i have not actully seen a statement one way or another on that | 16:18 |
| sean-k-mooney | the souce to image building for openstack sevices was pretty clear but amy didnt really say anything about the binary depencies fo them | 16:19 |
| andreykurilin | Hi folks! I have a question about recent Tempest version (46.1) and how it is supposed to work with pre-provisioned credentials. Some tests were updated to require `[“primary", "project_reader”]` credentials. However, when I define two credentials (one with the member role and another with the reader role), Tempest often selects the reader credential as primary. This then causes failures because Tempest cannot find another | 16:29 |
| andreykurilin | credential with the reader role. Some other tests also fail for a similar reason: the reader credential gets picked as primary, and the test then fails when attempting write operations. If I define two credentials with both member and reader roles, it effectively defeats the purpose of testing APIs with a dedicated reader role. | 16:29 |
| gmaan | andreykurilin: i think that might be bug in hashing logic in pre-provisioned creds, let me check the pre-creds job and how it is selecting the both | 17:34 |
| gmaan | andreykurilin: can you add your generated file in bug or if it is failing in upstream job, link to that will help | 17:34 |
| gmaan | andreykurilin: i saw tempest-full-test-account-py3 is brokemn for same reason. it was non voting and already broken for other reason so we did not notice it. | 17:46 |
| gmaan | anyways i logged the bug https://bugs.launchpad.net/tempest/+bug/2143564 and will start working on it | 17:47 |
| andreykurilin | gmann: oh, thank you! | 17:49 |
| andreykurilin | gmann: another question if you are still around. This project_reader change uncovered another issue - an example of acounts.yaml has domain and system scoped credentials. It also includes “regular” one. The regular one still uses “tenant_name” (and this is what we were using). project_reader credential tries to obtain credential with “project” scope. Credential specified with tenant_name (instead of project_name) is | 19:09 |
| andreykurilin | not treated as project scoped based on https://github.com/openstack/tempest/blob/master/tempest/lib/common/preprov_creds.py#L131 . I understand that everyone should move to project naming, but tenant_* is defined as still supported alias | 19:09 |
| gmaan | account_generator.py command dunmp the project_name in account file and that is what we have in CI jobs also - https://github.com/openstack/tempest/blob/cbb2eff6c75b190a21705abeb506420b26417b51/tempest/cmd/account_generator.py#L188 | 19:15 |
| gmaan | 'regular' you mean sample files right? | 19:15 |
| gmaan | this one https://github.com/openstack/tempest/blob/master/etc/accounts.yaml.sample | 19:15 |
| gmaan | i can fix that, we might have just forgot to update it | 19:16 |
| gmaan | but I will say those sample file can be outdated anytime and it is good to generate the file using tempest account-generator | 19:18 |
| opendevreview | Ghanshyam Maan proposed openstack/tempest master: Add tempest plugin testing in stable release testing https://review.opendev.org/c/openstack/tempest/+/978960 | 19:19 |
| andreykurilin | gmann: yeah, I understand that samples were out-dated, but the question is if tenant_name is an alias for project_name (and it works for auth and everything except project_reader), should it be also be an alias while making a decision about “scope”? | 19:31 |
| andreykurilin | https://github.com/openstack/tempest/blob/master/etc/accounts.yaml.sample yes, this one | 19:32 |
| gmaan | yeah, i can add alias, but that does not solve the problem, as tempest-full-test-account-py job failing even it has the project_name | 19:33 |
| andreykurilin | It is separate things :) just things that I found while trying to figure out why out automation does not work with latest version of tempest (we used to do not pin the version and rely on latest) | 19:34 |
| gmaan | I think it is having the same issue i fixed for dynamic creds, project reader and project members are created under separate projects instead of same project. Test create resources under one project and resder role try to access from other project | 19:34 |
| gmaan | I think continue supporting tenant_name is not an issue | 19:35 |
| andreykurilin | Btw, right now primary does not require any applied roles. Maybe the solution is just to make it require member role? Or “member” is not standartized across different clouds? | 19:36 |
| gmaan | ah, that also i fixed in dynamic creds but pending in pre-provosion creds | 19:38 |
| gmaan | will fix that also. | 19:38 |
| gmaan | lot of things to solve for pre-prov creds, I think I should fix all and make job voiting so that we can fix both dynamic as well as pre-prov | 19:39 |
| gmaan | for future issue | 19:39 |
| andreykurilin | Need any help? | 19:40 |
| gmaan | let me fix and propose it. will updatre you for testiing | 19:40 |
| andreykurilin | Thank you | 19:40 |
| gmaan | np! thanks for reporting | 19:41 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!